mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-06-01 16:11:14 -05:00
d965712a72
🔧 TYPE SYSTEM ENHANCEMENT: - Enhanced dtype parameter to accept Union[str, np.dtype, type] - Comprehensive type handling with proper error messages - Backward compatibility maintained 🧠 MEMORY LAYOUT ANALYSIS: - Added stride analysis and contiguous memory checking - Enhanced memory profiling with cache efficiency insights - New properties: strides, is_contiguous 📐 VIEW/COPY SEMANTICS: - Implemented view(), clone(), contiguous() methods - PyTorch-compatible memory sharing behavior - Proper gradient tracking preservation 🎯 IMPROVED ASSESSMENT QUESTIONS: - Replaced arithmetic with systems thinking questions - Focus on memory layout, broadcasting, and tensor operations - Grounded in actual student implementations ⚡ BROADCASTING ENHANCEMENTS: - Added comprehensive failure case demonstrations - Clear explanations of broadcasting rules - Production-relevant debugging insights All changes maintain educational clarity while adding technical depth that transfers directly to PyTorch/TensorFlow frameworks.